Optimization and data mining for shock-induced mixing enhancement inside scramjet using stochastic deep-learning flowfield prediction

Abstract

One of the significant challenges in supersonic combustion ramjet engines lies in effective mixing of the fuel, primarily due to the high momentum of supersonic inflow at the combustor. Among the various fluid phenomena associated with fuel mixing, it is well recognized that the interaction between the fuel-injection jet plume and oblique shock waves can significantly enhance mixing efficiency. However, the most suitable interaction for optimal mixing enhancement remains yet to be clarified. The present study conducts model-based optimization and data mining for shock-induced mixing enhancement of angled-slot injection in a two-dimensional scramjet combustor. Stochastic deep-learning flowfield prediction has been utilized to enable fast and reliable evaluations of a substantial number of designs. Prediction errors can be estimated without requiring correct data owing to uncertainty quantification techniques. Data mining and sensitivity analysis, coupled with flowfield prediction, have revealed the optimal shock interaction with the fuel jet plume characterized by a pronounced downstream recirculation region. The mechanism that drives mixing enhancement through this recirculation region has been discussed based on the results of optimization and sensitivity analysis. This study has yielded valuable insights for the future design of scramjet injectors. Furthermore, the effectiveness of the model-based design and analysis has been demonstrated through the present study, showcasing its potential for guiding future developments in scramjet technology.